Estimation of muscle force from mechanomyography (MMG) for physical human-robot interaction

Abstract

Mechaomyography (MMG) is the muscle surface oscillations generated by the dimensional change of the contracting muscle fibers. Since MMG reflects the number of recruited motor units (MUs) and their firing rates as electromyography (EMG) does, it can be used to estimate the exerted force by skeletal muscles. The aim of this study was to demonstrate the feasibility of MMG for estimating elbow flexion force at the wrist under an isometric contraction using an artificial neural network (ANN). We performed experiments with five subjects, and force at the wrist and MMG from three contributing muscles were recorded. It was found that MMG can be utilized to accurately estimate isometric elbow flexion force based on the values of normalized root mean square error (NRMSE=0.127$\plusmn0.015$) and the cross-correlation coefficient (CORR = 0.903$\plusmn0.025$). The estimation performance of MMG was evaluated in comparison with that of EMG under the same experimental condition. These experimental results suggest that MMG has potential for estimating muscle force, and its possible applications include physical human-robot interaction (pHRI) such as external prosthesis and exoskeleton robots.